toward better summaries

diff --git a/src/language/stats/ctables.c b/src/language/stats/ctables.c
index bef413a879d65d19aba3e3e546ae19ad42f00299..22ca909d9897d933b246a75e4bcb920c53479a11 100644 (file)
--- a/src/language/stats/ctables.c
+++ b/src/language/stats/ctables.c
@@ -224,9 +224,12 @@ enum ctables_label_position
 
 struct var_array
   {
-    const struct ctables_axis *summary;
     struct variable **vars;
     size_t n;
+
+    struct ctables_summary_spec *summaries;
+    size_t n_summaries;
+    struct variable *summary_var;
   };
 
 struct var_array2
@@ -239,6 +242,7 @@ struct ctables_table
   {
     struct ctables_axis *axes[PIVOT_N_AXES];
     struct var_array2 vaas[PIVOT_N_AXES];
+    enum pivot_axis_type summary_axis;
     struct hmap ft;
 
     enum pivot_axis_type slabels_position;
@@ -616,6 +620,21 @@ ctables_summary_default_format (enum ctables_summary_function function,
     }
 }
 
+static char *
+ctables_summary_default_label (enum ctables_summary_function function,
+                               double percentile)
+{
+  static const char *default_labels[] = {
+#define S(ENUM, NAME, LABEL, FORMAT, AVAILABILITY) [ENUM] = LABEL,
+    SUMMARIES
+#undef S
+  };
+
+  return (function == CTSF_PTILE
+          ? xasprintf (_("Percentile %.2f"), percentile)
+          : xstrdup (gettext (default_labels[function])));
+}
+
 static const char *
 ctables_summary_function_name (enum ctables_summary_function function)
 {
@@ -792,17 +811,8 @@ ctables_axis_parse_postfix (struct ctables_axis_parse_ctx *ctx)
           label = ss_xstrdup (lex_tokss (ctx->lexer));
           lex_get (ctx->lexer);
         }
-      else if (function == CTSF_PTILE)
-        label = xasprintf (_("Percentile %.2f"), percentile);
       else
-        {
-          static const char *default_labels[] = {
-#define S(ENUM, NAME, LABEL, FORMAT, AVAILABILITY) [ENUM] = LABEL,
-            SUMMARIES
-#undef S
-          };
-          label = xstrdup (gettext (default_labels[function]));
-        }
+        label = ctables_summary_default_label (function, percentile);
 
       /* Parse format. */
       struct fmt_spec format;
@@ -1299,11 +1309,19 @@ nest_fts (struct var_array2 va0, struct var_array2 va1)
           vars[n++] = b->vars[k];
         assert (n == allocate);
 
-        assert (!(a->summary && b->summary));
+        const struct var_array *summary_src;
+        if (!a->summary_var)
+          summary_src = b;
+        else if (!b->summary_var)
+          summary_src = a;
+        else
+          NOT_REACHED ();
         vaa.vas[vaa.n++] = (struct var_array) {
-          .summary = a->summary ? a->summary : b->summary,
           .vars = vars,
-          .n = n
+          .n = n,
+          .summaries = summary_src->summaries,
+          .n_summaries = summary_src->n_summaries,
+          .summary_var = summary_src->summary_var,
         };
       }
   var_array2_uninit (&va0);
@@ -1346,7 +1364,12 @@ enumerate_fts (enum pivot_axis_type axis_type, const struct ctables_axis *a)
           *axes = axis_type;
           *va = (struct var_array) { .vars = vars, .n = 1 };
         }
-      va->summary = a->scale || a->n_summaries ? a : NULL;
+      if (a->n_summaries || a->scale)
+        {
+          va->summaries = a->summaries;
+          va->n_summaries = a->n_summaries;
+          va->summary_var = a->var.var;
+        }
       return (struct var_array2) { .vas = va, .n = 1 };
 
     case CTAO_STACK:
@@ -1951,6 +1974,28 @@ ctables_execute (struct dataset *ds, struct ctables *ct)
             *va = (struct var_array) { .n = 0 };
             t->vaas[a] = (struct var_array2) { .vas = va, .n = 1 };
           }
+
+      for (size_t i = 0; i < t->vaas[t->summary_axis].n; i++)
+        {
+          struct var_array *va = &t->vaas[t->summary_axis].vas[i];
+          if (!va->n_summaries)
+            {
+              va->summaries = xmalloc (sizeof *va->summaries);
+              va->n_summaries = 1;
+
+              enum ctables_summary_function function
+                = va->summary_var ? CTSF_MEAN : CTSF_COUNT;
+              struct ctables_var var = { .is_mrset = false, .var = va->summary_var };
+
+              *va->summaries = (struct ctables_summary_spec) {
+                .function = function,
+                .format = ctables_summary_default_format (function, &var),
+                .label = ctables_summary_default_label (function, 0),
+              };
+              if (!va->summary_var)
+                va->summary_var = va->vars[0];
+            }
+        }
     }
 
   struct casereader *input = casereader_create_filter_weight (proc_open (ds),
@@ -2363,26 +2408,57 @@ cmd_ctables (struct lexer *lexer, struct dataset *ds)
 
       const struct ctables_axis *scales[PIVOT_N_AXES];
       size_t n_scales = 0;
-      for (size_t i = 0; i < 3; i++)
+      for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
         {
-          scales[i] = find_scale (t->axes[i]);
-          if (scales[i])
+          scales[a] = find_scale (t->axes[a]);
+          if (scales[a])
             n_scales++;
         }
       if (n_scales > 1)
         {
-          msg (SE, _("Scale variables may appear only on one dimension."));
+          msg (SE, _("Scale variables may appear only on one axis."));
           if (scales[PIVOT_AXIS_ROW])
             msg_at (SN, scales[PIVOT_AXIS_ROW]->loc,
-                    _("This scale variable appears in the rows dimension."));
+                    _("This scale variable appears on the rows axis."));
           if (scales[PIVOT_AXIS_COLUMN])
             msg_at (SN, scales[PIVOT_AXIS_COLUMN]->loc,
-                    _("This scale variable appears in the columns dimension."));
+                    _("This scale variable appears on the columns axis."));
           if (scales[PIVOT_AXIS_LAYER])
             msg_at (SN, scales[PIVOT_AXIS_LAYER]->loc,
-                    _("This scale variable appears in the layer dimension."));
+                    _("This scale variable appears on the layer axis."));
+          goto error;
+        }
+
+      const struct ctables_axis *summaries[PIVOT_N_AXES];
+      size_t n_summaries = 0;
+      for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+        {
+          summaries[a] = (scales[a]
+                          ? scales[a]
+                          : find_categorical_summary_spec (t->axes[a]));
+          if (summaries[a])
+            n_summaries++;
+        }
+      if (n_summaries > 1)
+        {
+          msg (SE, _("Summaries may appear only on one axis."));
+          if (summaries[PIVOT_AXIS_ROW])
+            msg_at (SN, summaries[PIVOT_AXIS_ROW]->loc,
+                    _("This variable on the rows axis has a summary."));
+          if (summaries[PIVOT_AXIS_COLUMN])
+            msg_at (SN, summaries[PIVOT_AXIS_COLUMN]->loc,
+                    _("This variable on the columns axis has a summary."));
+          if (summaries[PIVOT_AXIS_LAYER])
+            msg_at (SN, summaries[PIVOT_AXIS_LAYER]->loc,
+                    _("This variable on the layers axis has a summary."));
           goto error;
         }
+      for (enum pivot_axis_type a = 0; a < PIVOT_N_AXES; a++)
+        if (n_summaries ? summaries[a] : t->axes[a])
+          {
+            t->summary_axis = a;
+            break;
+          }
 
       if (lex_token (lexer) == T_ENDCMD)
         break;